Search Mailing List Archives
[protege-owl] Use of multiple necessary and sufficient blocks....and tree representation of equivalents
amarch at conceptum.com.ar
Mon Apr 23 12:16:03 PDT 2007
OK. Got it.
My confusion arose from the fact that, on detecting equivalences, Protégé
not only places them alonside the original class (as greyed classes) but
also -graphically- subclasses them. I wrote about this behaviour some time
ago, as I cannot quite understand why Protégé designers did this. Is there
a reason for this, or am I again out on a limb? It seems somehow
counterintuitive to have a class resulting both as an equivalent and a
subclass after running a classification. The problem arises when after
classification both real subclasses and equivalents are rendered as
(graphical) subclasses. They get mixed up and the final graphical
representation is difficult to read. Wouldn't it be more convenient, for
user's sake, to just leave equivalents greyed and on the same line as the
"main" class and linked by the Ξ sign?
Regarding Thomas' answer seconds ago, my case was something on the lines of
what he exemplifies in the last paragraph.
> -----Original Message-----
> From: protege-owl-bounces at lists.stanford.edu
> [mailto:protege-owl-bounces at lists.stanford.edu] On Behalf Of
> Thomas Russ
> Sent: Monday, April 23, 2007 3:48 PM
> To: User support for the Protege-OWL editor
> Subject: Re: [protege-owl] OWL individuals
> On Apr 21, 2007, at 10:02 AM, Alan March wrote:
> > Hi.
> > Sometime ago I posted a message inquiring as to the intended use of
> > multiple Necessary and Sufficient blocks. I did receive
> some answers,
> > but they were mostly examples, and did not quite answer my question.
> > So again, when should multiple N&S blocks be used? As far as I have
> > reasoned, they should be used when different groups of
> axioms, each of
> > them by themselves or together, allow for a complete
> definition of a
> > class.
> > Horridge et al's Owl Tutorial carries an example of *how*
> to establish
> > multiple N&S blocks, but, at least to the best of my
> undestanding, not
> > *when* to use them. As far as I can gather, it would seem that they
> > must be used in the manner I explained above. Thus, an
> individual may
> > be considered a member of the "triangle" class when it
> *either* "has
> > three angles and is a subclass of shape" **or** "has three
> sides and
> > is a subclass of shape", or *both*. When I emphasize
> "both", I mean to
> > say that if such individual fullfilled both N&S blocks, it
> would also
> > be a member of the triangle class.
> > So, my conclusion is that multiple N&S blocks would seem to
> boil down
> > to a sort of "and/or" situation, where a class may be
> defined as such
> > if it carries either block or both blocks. Am I right in this
> > assumption?
> Conceptually, I like to think of the necessary and sufficient
> conditions separately.
> There are some examples one can come up with where
> sufficient, but not necessary conditions apply.
> By separating the necessary and sufficient, one can then make
> a bit more sense of things like the triangle case.
> A triangle has 3 sides and 3 angles as necessary conditions.
> In other words, every triangle must have both 3 sides and 3 angles.
> Having 3 sides is a sufficient condition for being a triangle.
> Having 3 angles is a (separate) sufficient condition for
> being a triangle.
> The reason I like to separate such concerns has to do with
> the way inference works. If all that you know about a
> polygon is that it has
> 3 sides, then that is SUFFICIENT information to conclude it
> is a triangle. At that point, the necessary condition that
> it also have 3 angles comes into play.
> As an example of sufficient but not necessary conditions,
> consider the following conditions for being a Student:
> enrolled-in some University
> enrolled-in some Technical-Institute
> enrolled-in some Community-College
> The modeling of sufficient, but not necessary conditions in OWL (and
> Protégé) can be done by subclassing and putting both
> necessary and sufficient conditions on the subclass. Proper
> subclasses are sufficient but not necessary conditions for
> membership in their parent class.
> That generally means that most such issues in description
> logics are solved using the class/subclass system. In the
> student example, a common modeling method would be to have an
> umbrella concept like Institution-of-Higher-Education that
> subsumes all of the types listed, and then have a single N&S
> condition for that instead.
> But one can imagine situations where the individual
> conditions do not warrant having their own class definition,
> in which case, separate sufficient conditions may be justified.
> protege-owl mailing list
> protege-owl at lists.stanford.edu
> Instructions for unsubscribing:
More information about the protege-owl